Power amplifiers in wideband radio systems are often used to amplify wideband signals or signal combinations with high peak to average power ratio, PAR. The amplifiers must then be able to repeatedly output very high power for very short periods, even though the bulk of the output power is generated at the much lower average power level. In systems with random phase combinations of many signals (without any dominating ones) the amplitude of the signal follows a Rayleigh distribution.
A conventional single-transistor power amplifier (for example a class B, AB or F power amplifier) has a fixed radio frequency (RF) load resistance and a fixed voltage supply. The bias in class B or AB amplifiers causes the output current to have a form close to that of a pulse train of half wave rectified sinusoid current pulses. The direct current (DC) current (and hence DC power) is therefore largely proportional to the RF output current amplitude (and voltage). The output power, however, is proportional to the RF output current squared. The efficiency, i.e. output power divided by DC power, is therefore also proportional to the output amplitude. The average efficiency of a power amplifier is consequently low when amplifying signals that on average have a low output amplitude (or power) compared to the maximum required output amplitude (or power), i.e. high PAR.
An example of a Doherty amplifier is described in “A new high efficiency power amplifier for modulated waves,”, W. H. Doherty, Proc. IRE, vol. 24, no. 9, pp. 1163-1182, September 1936. Doherty amplifiers such as this have high average efficiency for amplitude-modulated signals with high peak-to-average ratio (PAR) since they have a much lower average sum of RF output current magnitudes from the transistors at low amplitudes. This causes high average efficiency since the DC currents drawn by the transistors are largely proportional to the RF current magnitudes.
Reduced RF output currents are obtained by having high transimpedance from at least one transistor to the output, while having the possibility of in-phase combining all transistor outputs to obtain full output power. Higher transimpedance means higher voltage at the output for the same amount of current. This is achieved in the Doherty amplifier by having the main transistor (“carrier amplifier”) displaced from the output node by a quarter wavelength transmission line of characteristic impedance Ropt, (where a transistor's Ropt is the optimal load resistance for achieving maximum output power).
Since the load Rload has a lower value than Ropt (typically Rload=Ropt/2) this line acts as a quarterwave transformer. The transimpedance to the output from the main transistor is equal to the characteristic impedance of the quarterwave line (i.e. Ropt), instead of Rload as would be the case for one transistor coupled directly to the load. The self-impedance at the main transistor is increased quadratically to the characteristic impedance squared divided by Rload (aka “impedance inversion” of the load). If the peak transistor (also known as “auxiliary amplifier” or “peaking amplifier”) has an Ropt that in parallel combination with the Ropt of the main transistor gives Rload, full combined output power will be possible by in-phase combining (i.e. adjusting the phase (time, electrical length) difference between the main and peak drive signals so the output waves from both are in phase at the output Rload).
The carrier amplifier output current is linear in amplitude, i.e. follows the desired output signal. The peaking amplifier output current is zero for low amplitudes, and rises (piecewise) linearly from the transition point. The transition point for a 2-stage Doherty designed for two equal size transistors is at half the maximum output amplitude. The shaping of the output RF current amplitude is in some cases done by biasing the gate low and increasing the RF drive voltage, known as class C operation. This shaping can also be done, wholly or partially, earlier in the processing chain, by analog or digital signal shaping circuits.
A first way to extend the Doherty amplifier to more stages (transistors, constituent amplifiers) was shown by F. H. Raab in a paper entitled “Efficiency of Doherty RF Power Amplifier Systems”, IEEE Trans. Broadcasting, vol. BC-33, no. 3, pp. 77-83, September 1987. These amplifiers can be described as having a cascade of quarterwave transmission lines with successively lower characteristic impedance towards the output (load), where RF transistors are connected at the junctions between the transmission lines. The resulting amplifier makes it possible to have high efficiency in a wider range of back off.
U.S. Pat. No. 8,022,760 discloses an alternative arrangement for 3-transistor Doherty amplifiers, whose main benefit is better placement of the transition points (corresponding to high points in the efficiency vs. amplitude curve) for equal-sized transistors. Higher order versions of the 3-transistor Doherty amplifier in U.S. Pat. No. 8,022,760 consist of having a higher order quarterwave cascade multistage Doherty as a peaking amplifier. Only the ones with an odd total number, N, of transistors (5, 7, 9 etc . . . ) work, i.e. those that have quarterwave cascades with an even number, N−1, of quarterwave lines.
EP2,403,135 discloses a four-transistor Doherty amplifier. This is basically the 3-stage amplifier of U.S. Pat. No. 8,022,760 with an added peaking amplifier at the output node and has largely the same advantages as U.S. Pat. No. 8,022,760 regarding transistor sizes. Higher order versions of EP2,403,135 consist of even numbers, N, of transistors, with both a directly connected and a quarterwave-connected transistor at the output node. The quarterwave cascade in the peaking amplifier branch will therefore have the total length, N−2, i.e. the same lengths as for the amplifiers in U.S. Pat. No. 8,022,760.
The multistage Doherty amplifiers by Raab generally have their transition points too high to give good average efficiency with high-PAR signals if the transistor stages are of equal size. FIGS. 1a, 1b and 1c show the curves for a 4-stage implementation, in which the lowest transition point is at 0.37 of full output. The amplifiers with higher numbers of stages generally have the same problem, as do arrangements in which a small number of different transistor sizes are available.
The requirement for several different amplitude-limited drive signals can pose a problem in some implementation technologies, for example increased implementation complexity.
Referring to FIGS. 2a to 2c, the 5-stage amplifier according to U.S. Pat. No. 8,022,760 has advantages over those of Raab for use with high-PAR signals, since the lowest transition point with five equal sized transistors is at 0.2 of full output amplitude (−14 dB). However, it has a sparse distribution of transition points at low amplitude.
For six and higher numbers of stages, implementations with equal size transistors of U.S. Pat. No. 8,022,760 and EP2,403,135 all suffer from too sparse transition points at low output amplitudes, as illustrated by FIGS. 3a to 3c for a 6-stage arrangement according to EP2,403,135 and FIGS. 4a to 4c for a 7-stage amplifier according to U.S. Pat. No. 8,022,760.
Thus, each of the amplifier arrangements described in FIGS. 2, 3 and 4 have the disadvantage of requiring several amplitude limited drive signals, and also have the disadvantage of having a poor distribution of transition points.